Vibe Coding Empowers Lawyers to Influence AI Winners, Losers

During his annual presentation in November on the state of artificial intelligence, tech analyst Benedict Evans explained how AI is “eating the world.” After attending last month’s LegalWeek in New York, I can say it doesn’t seem like the legal profession will be immune from that appetite.

“Vibe coding,” the practice of building software using natural language prompting, is championed by its evangelists as the future of software development, and its adoption has now spread to law firms and its partners.

Just a few weeks ago, markets priced in their own view of the potential of AI-first software development, when Anthropic PBC’s release of an automation plug-in for legal work wiped out billions in stocks. How this practice develops inside law firms raises important questions about the rapid progression of legal technology—and will have consequences for the business of law.

In a quarterly earnings call last year, Microsoft Corp. reported that 90% of the Fortune 100 were using its GitHub Copilot product for AI coding. An even more compelling stat: SemiAnalysis has projected that at the current trajectory, Anthropic’s Claude Code would account for over 20% of all daily code changes by the end of 2026.

In legal work, the integration of large language models into daily practice has been equally extensive. Data shows that 83% of lawyers in 2026 have “broad AI access,” which is up from 61% in 2025. This signals that the barrier to software development for nontechnical professionals has collapsed.

For lawyers, whose work is already largely text-based and increasingly AI-assisted, vibe coding is a natural next step. Junior associates and partners alike can now conceptualize a solution to a workflow problem and build a working prototype without manually writing a line of code themselves.

The immediate benefits are obvious. Lawyers who can harness vibe coding’s potential early are achieving significant productivity gains in a profession where competitive edges are measured in hours. At LegalWeek, several panels featured insights from partners and innovation teams who had built vibe-coded products with great success. Such examples should give more AI-skeptic partners within law firms food for thought.

The use of untested, unregulated software products in legal work could be the beginning of a cautionary tale. Many lawyers understand the specter of governance crises that vibe coding raises—and not without reason.

But outlawing vibe coding by lawyers would be an overcorrection and would miss the point of what this phenomenon means for the business of law.

Hundreds of legal-tech companies now exist whose value propositions blur together across document upload, drafting, diligence, and deal management. This indicates how buoyant the sector is but is also overwhelming for law firms at the point of procurement.

Last summer, the sector was so flush with cash that there was constant talk of a funding bubble in legal technology. But vibe coding tools (such as Claude Code) that produce software almost as well as the average software developer improved rapidly in just a few months, and law firms now can flip the script on how legal technology is evaluated.

Law firms must realize that vibe coding can be the most direct form of product research for legal technology ever produced. What lawyers design themselves—and what they don’t bother building—is a novel way of discovering an unmet need.

In a market many see as oversaturated, this new mechanism for evaluating technology may be what decides the winners and losers in a space that otherwise was left to competition between legal-tech startups that are inherently one rung removed from the lawyer-client relationship.

Partners should resist the temptation to view this as an “either-or” scenario in which law firms diverge on accepting internal vibe coding. First, law firms must consider what vibe coding means for capital allocation. Vibe coding output should be seen simply as a demand signal from their biggest assets: lawyers themselves.

Following these signals, they should concentrate their technology spend on the demonstrably valuable tools that emerge from this experimentation, rather than scattering budgets across dozens of overlapping vendors.

Second, law firms must learn to inhabit the economics of software markets. As has always been the case with enterprise software, successful tech experiments inside law firms will eventually end up being turned into products by external software vendors and then sold back to the firms themselves.

Any individual firm’s homegrown tool will be a temporary advantage at best. The basic reality of software economics—in which most of the cost sits in maintaining a tool over time rather than in the first version build—compounds this. Vibe-coded prototypes aren’t built to last, but they can be built to learn from.

The law firms that ban vibe coding will lose twice: once by missing feedback in the form of product intelligence from their lawyers, and again when they can’t tell what technology to build from what to buy. To seize this opportunity, partners must be open-minded to the concept of experimentation, and willing to invest in the architecture required to ensure vibe coding remains a strategic advantage and not a technical liability.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Author Information

Andrew Thompson is chief technology officer of Orbital.

Write for Us: Author Guidelines